Article pubs.acs.org/Macromolecules
A Journey along the Extruder with Polystyrene:C60 Nanocomposites: Convergence of Feeding Formulations into a Similar Nanomorphology Hugo Gaspar,† Paulo Teixeira,† Raquel Santos,† Liliana Fernandes,† Loic Hilliou,† Michael P. Weir,‡ Andrew J. Parnell,‡ Kerry J. Abrams,§ Christopher J. Hill,∥ Wim G. Bouwman,⊥ Steven R. Parnell,⊥ Stephen M. King,# Nigel Clarke,‡ José A. Covas,*,† and Gabriel Bernardo*,† †
Institute for Polymers and Composites/I3N, University of Minho, 4800-058 Guimarães, Portugal Department of Physics and Astronomy, The University of Sheffield, Sheffield S3 7RH, United Kingdom § Department of Materials Science and Engineering, The University of Sheffield, Sheffield S1 3JD, United Kingdom ∥ Department of Biomedical Science, The University of Sheffield, Sheffield S3 7HF, United Kingdom ⊥ Faculty of Applied Sciences, Delft University of Technology, Mekelweg 15, 2629 JB Delft, Netherlands # ISIS Pulsed Neutron Source, STFC Rutherford Appleton Laboratory, Harwell Campus, Didcot OX11 0QX, United Kingdom ‡
S Supporting Information *
ABSTRACT: We investigated the effect of the feeding formulation (premixed powders of pure components versus solvent-blended mixture) of polystyrene−C60 composites on the dispersion and reagglomeration phenomena developing along the barrel of a twinscrew extruder. The dispersion of C60 in the PS matrix is studied over different length scales using a combination of optical microscopy, spinecho small-angle neutron scattering (SESANS), small-angle neutron scattering (SANS), small-angle X-ray scattering (SAXS), and wideangle X-ray scattering (WAXS). When a solvent-blended mixture is used as the feeding formulation, the inlet material contains essentially molecularly dispersed C60 as revealed by the nanodomains with very small phase contrast. However, C60 reagglomeration occurs along the extruder, creating a morphology still containing only nanodomains but with much higher phase contrast. In the case of mixed powders, the material evolves from the initial macroscopic mixture of pure polystyrene and C60 into a composite simultaneously containing micro- and nanoaggregates of C60 as well as C60 molecularly dispersed in the matrix. Our results show that the two different initial feeding formulations with widely different initial morphologies converge along the extruder, through opposite morphological pathways, into a similar final nanomorphology which is dictated by the interplay between the thermodynamics of the system and the flow. Correlations between the morphological evolution along the extruder and the thermorheological properties of the composites are identified.
1. INTRODUCTION
control, with both thermodynamic and kinetic processes playing significant roles. The dispersion of carbon nanotubes and graphenes is complicated by their strong van der Waals interactions, chemical inertia that creates weak interfaces with most polymers, and difficulty to control initial size and shape of these nanoparticles. These parameters influence the cohesiveness of the agglomerates and have been related to dispersion routes such as rupture and erosion.12,13 Conversely, fullerene C60 (also known as buckminsterfullerene) has a very welldefined size and shape: it is a bucky-ball with a diameter of 0.7 nm.14,15 When replacing graphenes or carbon nanotubes by
Polymer nanocomposites containing carbon nanoparticles such as graphenes,1,2 carbon nanotubes,3−7 and fullerenes8 open a new horizon for polymeric-based materials. These materials benefit from a synergistic combination of useful polymer properties, such as low density, flexibility, ease of processing, and cost efficiency, with nanoparticles that not only significantly enhance the mechanical, electrical, and thermal properties of the resulting composite9 but also create novel functionalities. Although the homogeneous dispersion and distribution of carbon nanoparticles in a polymer matrix is generally desired to maximize performance, some level of particle aggregation can be advantageous in specific applications.10,11 Though property improvements have been achieved in a variety of nanocomposites, carbon nanoparticle dispersion remains difficult to © XXXX American Chemical Society
Received: October 21, 2016 Revised: April 4, 2017
A
DOI: 10.1021/acs.macromol.6b02283 Macromolecules XXXX, XXX, XXX−XXX
Article
Macromolecules
intermeshing corotating twin-screw extruder. We study and compare two different types of melt-processed PS−C 60 composites that were prepared using identical C60 loadings (1 wt %) and processing conditions (flow rate = 130 g h−1, screw speed = 80 rpm; extruder and die set temperature = 200 °C), but which differ widely on the quality of dispersion of the initial material: (i) mixed powders of pure polystyrene and pure C60 and (ii) a solvent preprocessed PS-C60 blend. The morphological characterization of the composites was performed using both direct and indirect characterization techniques. Five direct morphological characterization techniques were used to probe four different length-scale regimes, namely, optical microscopy for the ∼1 mm−1 μm range; spin-echo small-angle neutron scattering (SESANS)35 for the ∼20 μm−200 nm range; smallangle neutron scattering (SANS) and small-angle X-ray scattering (SAXS) for the ∼200 nm−1 nm range; and wideangle X-ray scattering (WAXS) for the subnanometer range. Additionally, the composites were characterized by differential scanning calorimetry (DSC) and rheological measurements, as indirect morphological characterization techniques, to provide an overall understanding of the processing−structure−property relationships in these model polymer−fullerene systems. This paper is organized as follows. We first describe the sample preparation methods. Then we characterize the dispersion of C60 in the composite samples using optical microscopy, SESANS, SANS, SAXS, and WAXS. Next, we study the variation of the glass transition temperature with sample preparation conditions, and we characterize the samples rheologically. Finally, we discuss and rationalize all the results obtained.
fullerenes in a polymer composite, particle size and geometry change (these being associated with entropic changes) while the differences in enthalphic interactions between the particles and the matrix are kept at a minimum. For this reason, polymer composites with C60 can be used as model systems to help understand the dispersion of carbon nanoparticles. In the past two decades, polystyrene (PS) composites containing fullerene nanoparticles have been the subject of a number of studies15−33 reporting improvements in properties such as electrical conductivity,22,23 thermal stability,24−27 and permeability.28,29 The impact of C60 loading and sample preparation conditions on the corresponding glass transition temperature (Tg) and melt viscosity has also received attention.16,19,34 Campbell et al.21 used wide-angle X-ray scattering (WAXS) to determine the miscibility limit of C60 in PS as being ∼1 wt %. In this context miscibility is understood as an absolute thermodynamic value associated with molecular homogeneity and which is reversible and independent of the pathway followed from aggregates to molecular dispersion; i.e., it is a maximum value that is independent of processing. An identical miscibility threshold was determined by Sanz et al.34 using a combination of microscopy, SANS, and WAXS experiments. Additionally, these authors also determined a dispersibility threshold of ∼4 wt % C60, where this is the maximum C60 loading associated with the maximum observed increase in the Tg of the composites. Loadings beyond this concentration gradually reverted the composite Tg toward the neat PS value. Increases in Tg with C60 loadings in PS composites were also reported by Weng et al.32 and by Wong et al.,19 with the latter also concluding that beyond 1 wt % nanoparticle concentration the C60 is aggregated in polydispersed aggregates. Self-assembly of C60 into clusters within PS−C60 thin films was also reported, with the size of the clusters becoming macroscopic for C60 concentrations in the range 3−4 wt % C60.30 The impact of C60 on the melt viscosity of PS−C60 composites was studied by Tuteja et al.16 They showed that a viscosity reduction in the PS−C60 nanocomposites may occur if the polymer is entangled (Mw > Mc, where Mc is the critical molecular mass for entanglement coupling) and if the average interparticle half-gap (h) is less than the polymer size (i.e., h < Rg), where h/a = [Φm/Φ]1/3 − 1, Φm being the maximum random packing volume fraction (∼0.638) and a being the particle radius (∼0.4 nm). For nonentangled polymer melts (Mw < Mc), a viscosity increase is observed upon the addition of C60. In practice, the authors observed a viscosity decrease in well-dispersed fullerene−polystyrene nanocomposites prepared via rapid precipitation from solution and using a PS with Mw = 393 kDa. On the other hand, a viscosity increase was observed when using a PS with Mw = 19.3 kDa. In most of the preceding studies, PS−C60 blending was achieved through the use of solvents, using methods such as solvent casting or rapid precipitation from solution. Despite the fact that in the commodity plastics industry melt extrusion is a ubiquitous process for the manufacturing of polymer nanocomposites, with solvents rarely being used, in the scarce literature studies in which polymer−C60 blending was performed in an extruder no attempts were made to assess the dispersion achieved.27 Therefore, there is an obvious fundamental and practical interest in studying how C60 nanoparticles can be mixed with polystyrene in a melt extruder. In this work, we conduct a detailed investigation of the dispersion of fullerene C60 in polystyrene composites along an
2. EXPERIMENTAL SECTION 2.1. Materials. The C60 fullerene used in this work was supplied by Solenne BV (>99.5% purity, Mw = 720.64 g mol−1 and density = 1.65 g cm−3). The polystyrene was purchased from Sigma-Aldrich (catalog # 430102) with average Mw ∼ 192 000, MFI = 6.0−9.0 g/10 min (200 °C/5 kg), and density of 1.05 g cm−3 at 25 °C. The polymer Rg is approximately 12 times larger (estimated from Rg ≅ 0.27Mw1/2 36) than the diameter of C60 (∼1 nm). 2.2. Nanocomposite Preparation. Two different types of PS− C60 formulations were prepared as feeding material for the twin-screw extruder: (a) 1.0 wt % C60 was simply mechanically mixed with PS (hereafter simply referred to as “mixed powders”), and (b) mixtures with 1.0 wt % C60 were prepared by solution blending followed by coprecipitation in a nonsolvent (hereafter referred to as “pre-solvent blended”). In the preparation of the mixed powders formulation, the granulated polystyrene was milled into a fine powder and then thoroughly mixed with the, as purchased, fine C60 powder in a rotating mixer. In the preparation of the pre-solvent blended formulation, appropriate amounts of PS and C60 were dissolved in toluene, while stirring at 80 °C for 4 h. Then, the solution was added dropwise into a 5-fold volume excess of precooled methanol (T < 0 °C) under continuous stirring, to enable the coprecipitation of the PS−C60 composite. The solids were vacuum filtered from methanol using a nylon membrane filter with a pore size of 0.45 μm (Whatman, cat no. 7404-004) and washed with cold methanol. This was followed by drying at ∼100 °C and 10−2 mbar for several hours, until no mass changes were detected using a balance with a precision of ±0.01 g. Upon drying, light purple fiber composites were obtained. The yield of composite preparation was ∼100%, showing that no C60 and polystyrene losses had occurred. The total amount of pre-solvent blended formulation prepared was ∼35 g (after drying). Finally, the fibrous composite was milled into a fine light purple powder. The two types of feeding formulations were then melt processed in a prototype corotating intermeshing twin-screw extruder of modular B
DOI: 10.1021/acs.macromol.6b02283 Macromolecules XXXX, XXX, XXX−XXX
Article
Macromolecules construction (Figure S.I.1.(a) in Supporting Information), designed to process small amounts of material (in the range of 30−300 g/h) while retaining the same characteristics of larger equipment. The screws have a diameter of 13 mm and an L/D ratio of 27 (Figure S.I.1.(b)). The screw profile is built by sliding along a shaft conveying (with distinct pitches and lengths) and kneading elements (that can be stacked with variable staggering angles to induce different hydrodynamic stress levels and therefore variable balances of distributive/dispersive mixing). The barrel contains nine material sampling ports, evenly distributed along its length. These are manually operated rotary valves which allow one to quickly remove small volumes of material from within the extruder during steady state operation.37,38 A miniaturized prototype volumetric feeder is positioned upstream. The screw profile used in this work contained two kneading zones consisting of four 3 mm thick kneading disks staggered at −30° (for intensive mixing) separated by a conveying zone. The flow channels had a maximum depth of 1.5 mm so that shear rates were relatively high, even at low screw speeds. Each PS−C60 mixture was compounded at a flow rate of 130 g/h, with the screws rotating at 80 rpm. The temperature profile along the barrel and the slit die (5 mm × 1 mm) was kept at a constant 200 °C for all the processing experiments. In order to track the quality of the mixing along the extruder, spherically shaped composite samples in their molten state were collected at sampling ports P2 and P9 during steady state extruder operation. These molten samples were then sandwiched between two Teflon sheets placed between two metal plates and slightly compressed, while cooling down to room temperature, into approximately circular disks with thickness ∼1.5 mm and diameter ∼20 mm (see Figure S.I.2). These samples, hereafter simply referred to as samples P2 and P9 depending on the sampling port from which they were collected, as well as the extruded ribbons collected after air cooling and winding were then stored for subsequent characterization. To keep the extrusion line running continuously, the lowest possible draw-down (i.e., the ratio between the linear velocity of the winder and the extrudate velocity) was applied. Additionally, reference samples with C60:PS weight ratios 1:99 were prepared by compression molding: (i) a mixture of the pure powders at 170 °C and 10 MPa during 5 min (one sample produced); (ii) powders of the solvent blended composite at 90 and 170 °C and 10 MPa during 5 min (two samples produced). These three reference samples were also stored for subsequent characterization. All the samples are shown in Figure S.I.2. The initial sample compression molded from mixed powders is mostly colorless and transparent, albeit containing a few macroscopic black spots in it. The initial pre-solvent blended sample is purple and transparent. The samples from mixed powders collected at P2, P9, and ribbon are all dark brown and optically opaque. The samples from pre-solvent blend collected at P2, P9, and ribbon are light brownish and still optically transparent to some degree. 2.3. Nanocomposites Characterization. 2.3.1. Optical Microscopy. The agglomerate size distribution in the initial, i.e., before melt compounding, pre-solvent blended and mixed powder formulations was determined by light transmission optical microscopy (OM) using a BH2 Olympus microscope coupled to a Leica DFC 280 camera, with a 1.6× ocular and 20× objective magnification. For microscopy observations, the feeding formulations (in powder form) were softened at 160 °C and spread onto a glass slide. The evolution of dispersion along the extruder was estimated by OM. For this, 10 μm thin sections were cut from samples at room temperature with a Leitz 1401 microtome using glass knifes with an angle of 45°. The extruded tapes were molded in an epoxy resin and cut perpendicular to the flow direction. Micrographs were acquired using the same microscopy equipment and under similar magnifications. To obtain sufficient statistics, at least six micrographs were analyzed using ImageJ Software, leading to an investigated total area of 2.1 mm2. The level of dispersion was quantified in terms of area ratio (Ar), which balances the total area of agglomerates and the total area analyzed, and cumulative relative distribution of the agglomerates
(CC60). This is determined by summing the areas of the individual agglomerates in ascending area order and dividing by the total area of agglomerates. In particular, the size of the larger agglomerate contained in 75% (Aχ75%) and 90% (Aχ90%) of the total area of agglomerates were followed. 2.3.2. Spin-Echo Small-Angle Neutron Scattering (SESANS). Spinecho small-angle neutron scattering (SESANS) is a relatively new technique to measure structures of materials. Structures can be determined over 3 orders of magnitude in length scale, from 10 nm to 20 μm. This is 2 orders of magnitude larger than conventional SANS and comparable to what may be studied with light scattering or OM, though with the complementary benefits afforded by the use of neutrons such as contrast and probing the bulk. However, like OM, and unlike SANS (or SAXS), SESANS measures in real space. Briefly, SESANS uses a series of magnetized permalloy films and magnetic fields to encode the scattering angle information in the spin precession of a beam of polarized neutrons. Neutrons scattered through different angles traverse magnetic fields of different lengths and thus precess differently to the unscattered neutrons. The structural length scale probed depends upon the applied magnetic field strength, hence varying the magnetic field strength provides tunability over a range of length scales, termed the spin-echo length. The SESANS measurements reported here were performed at the Reactor Institute Delft (TU Delft, Netherlands), using a beam monochromatised (using a pyrolytic graphite crystal) to a wavelength λ = 2.06 Å.35 The pre-solvent blend sample was measured for 4 h while the sample from mixed powders was measured for ∼12 h over the same spin-echo length range. 2.3.3. Small-Angle Neutron Scattering (SANS). Small-angle neutron scattering (SANS) measurements were performed at the LOQ diffractometer at the ISIS Pulsed Neutron Source (Rutherford Appleton Laboratory, Oxfordshire, UK) with a polychromatic incident beam of λ = 2−10 Å and fixed sample-to-detector distance of 4 m, to provide a scattering vector range of 0.009 < q < 0.25 Å−1, where q = (4π/λ) sin(θ/2) and θ is the scattering angle. Because of the high carbon to hydrogen content in C60, there is a naturally high neutron scattering length density contrast with the hydrogenous polystyrene polymer removing the need for isotopic substitution (deuteration). However, there is a greater degree of incoherent background scattering from the matrix that degrades the signal-to-noise, particularly at larger q values. The samples P2 and P9 were ∼1.5 mm thick, and the extruded ribbons were ∼0.7 mm thick although their exact thicknesses were measured by a micrometer and those values used in the data reduction procedure to ensure proper intensity scaling. Samples were mounted on a computer-controlled sample changer and SANS patterns were recorded at room temperature for approximately 90 min/sample. Each raw scattering data set was then radially averaged, corrected for the detector efficiency, sample transmission, and background scattering, and then converted to scattering cross-section data and plotted on an absolute scale (∂Σ/∂Ω vs q) using Mantid software.39 For convenience, we shall follow the normal convention of referring to ∂Σ/∂Ω as intensity (I). The corrected data were then fitted to appropriate models using SasView software (Version 3.1.1).40 The neutron scattering length densities of hydrogenous PS and of C60 are respectively ρN(H-PS) = 1.41 × 10−6 Å−2 and ρN(C60) = 5.50 × 10−6 Å−2 and therefore ΔρN = 4.09 × 10−6 Å−2. 2.3.4. Small-Angle X-ray Scattering (SAXS). SAXS measurements were performed in-house on a Bruker instrument (NanoStar, Department of Chemistry, University of Sheffield, UK) equipped with a microfocus Cu Kα source (8 keV, λ = 1.54 Å; Xenocs, France), collimating system with motorized scatterless slits (Xenocs, France) and a HiStar 2D multiwire gas detector (Siemens/Bruker). Scattering patterns were corrected for the detector’s dark current, spatial distortion, flat field, and normalized using sample thickness, exposure time, sample transmission, and the detector normalization coefficient. The electron densities of hydrogenous PS and of C60 are respectively ρX(H-PS) = 9.56 × 10−6 Å−2 and ρX(C60) = 1.40 × 10−5 Å−2 and therefore ΔρX = 4.44 × 10−6 Å−2. C
DOI: 10.1021/acs.macromol.6b02283 Macromolecules XXXX, XXX, XXX−XXX
Article
Macromolecules 2.3.5. Wide-Angle X-ray Scattering. WAXS measurements, of samples P9 from both mixed powders and solvent processed blend, were performed on a Xeuss 2.0 SAXS/WAXS laboratory beamline using a liquid Gallium MetalJet (Excillum) X-ray source (9.2 keV, λ = 1.34 Å). The scattered X-rays were detected using a Pilatus3R 1M detector. Scattering from the samples was collected at room temperature for 3 min. 2.3.6. Differential Scanning Calorimetry. Differential scanning calorimetry (DSC) experiments were performed on a PerkinElmer Pyris-1 calorimeter under nitrogen where temperature and heat capacity were calibrated using a sapphire standard. Thermal history was eliminated by ramping from 25 to 150 °C at 10 °C min−1, isothermal annealing at 150 °C for 2 min, and then cooling to 30 °C at 10 °C min−1. Two heating−cooling cycles were run. The glass transition temperature Tg was computed from the second heating run following the half-Cp extrapolation method. Error bars in Tg are estimated by the maximum deviation of three independent measurements. 2.3.7. Rheometry. Disks were loaded between the parallel plates (diameter 25 mm) of a stress-controlled rotational rheometer (ARG2, TA Instruments) at a temperature of 200 °C under air. Time was left for the samples to thermally stabilize and adjust to the defined gap between the parallel plates, as inferred from the online reading of the normal force relaxation. A time sweep was first performed at 1 Hz with a deformation of 0.1% (corresponding to the linear regime of viscoelasticity as inferred from separate strain sweep experiments performed at 1 Hz) to confirm the thermal stability of all samples within 20 min. Then a new sample was loaded and equilibrated as mentioned above, for the recording of a mechanical spectrum using a frequency sweep performed with a deformation of 0.1%.
The level of dispersion of C60 in the initial feeding formulations and in the samples collected along the extruder was first investigated by optical microscopy (OM), as shown in Figure 2 and Figure S.I.3. Figure S.I.3 shows optical microscopy images of the initial feeding formulations. The image of the “mixed powders” feeding formulation shows large black aggregates in a matrix of essentially pure polystyrene. Overall, this formulation shows both a poor dispersion and poor distribution of the aggregates. By contrast, on the image of the “solvent-blended” feeding formulation, although some randomly distributed darker spots are visible, these are not large black aggregates. These darker spots most likely result from some possible heterogeneities (regions with different PS:C 60 ratio) inherent to the precipitation process used in the preparation of this composite. Also visible in the “solvent-blended” figure is a larger, approximately circular, feature with ∼50 μm diameter which corresponds to material that did not melt during the preparation of the optical microscopy samples. The morphology development of PS nanocomposites containing 1 wt % of C60 along the extruder is presented in Figure 2a. As clearly shown, the extruded samples prepared from mixed powders contain a large number of microscopic aggregates. By contrast, in the extruded samples prepared from pre-solvent blend the number and size of microscopic features is much smaller. Figure 2b displays the corresponding evolution of the area ratio, and the dispersion characterization results are summarized in Table 1. As shown in Figure 2 and Table 1, in extruded samples from both mixed powders and pre-solvent blend, the area ratio remains approximately constant along the extruder. It is well accepted that dispersion of solid agglomerates in a molten matrix occurs when the hydrodynamic stresses developed during flow are larger than the cohesive strength of the agglomerates. This balance is usually quantified by the fragmentation number, Fa. It has been suggested that when Fa is large, dispersion is dominated by the rupture of the agglomerates into successively smaller aggregates. When Fa is small, erosion should prevail. This is a much slower dispersion route, whereby small aggregates or individual particles detach from the agglomerates. Both phenomena were reported in studies of the dispersion of carbon nanonotubes and graphite nanoplates.41,42 In the present work, as the area ratio remained constant along the extruder, this means that no substantial rupture took place, but some dispersion may have developed via erosion, as in this case the large agglomerates are still visible by optical microscopy. The SESANS data from two nanocomposites extracted from port P9 are shown in Figure 3. The data from the mixed powder sample have been fitted to a model of monodispersed particles using the Gaussian approximation formalism detailed by Andersson et al.,43 and the resulting fit is also shown in Figure 3. This modeling approach is used as the system is low concentration, and no consideration of a structure factor is necessary. The best fit to the data, in which the radius was the only free parameter, was found for particles of radius 2.83 ± 0.10 μm. The scattering length densities were fixed at the values quoted in Table 2, as derived from the SANS data analysis detailed below. We note in passing that the extended tail of the data at longer spin-echo lengths suggests that there is in fact a dispersion in the particle sizes of the agglomerates.
3. RESULTS AND DISCUSSION As an aid to the analysis and discussion of our experimental results, we start by showing in Figure 1 an idealized schematic of the different possible stages of C60 dispersion in a polymer matrix.
Figure 1. Idealized schematic of C60 dispersion in the polymer matrix along the extruder, starting from macroscopic agglomerates of C60 in a matrix of pure polymer and ending with an idealized mixture of individual C60 molecules homogeneously dispersed in the polymer matrix.
There are two extreme situations, namely situation A in which macroscopic and microscopic agglomerates of C60 coexist in a matrix of pure polymer and situation D consisting of an idealized mixture of individual C60 molecules homogeneously dispersed in the polymer matrix. Two intermediate cases (B and C) correspond to a mixture of microscopic and nanoscopic agglomerates of C60 in a polymer matrix containing some molecularly dispersed C60 and to a mixture of nanoscopic agglomerates in a polymer matrix containing a considerable amount of molecularly dispersed C60, respectively. D
DOI: 10.1021/acs.macromol.6b02283 Macromolecules XXXX, XXX, XXX−XXX
Article
Macromolecules
Figure 2. (a) Morphology development of PS nanocomposites containing 1 wt % of C60 along the twin-screw extruder. (b) Area ratio evolution along the twin-screw extruder of PS nanocomposites containing 1 wt % of C60.
Table 1. Optical Microscopy Dispersion Characterization Results for PS Nanocomposites Containing 1.0 wt % of C60 pre-solvent blended 2
2
from mixed powders 2
2
position
Aχ75% (μm )
Aχ90% (μm )
N (per mm )
Aχ75% (μm )
Aχ90% (μm2)
N (per mm2)
P2 P9 ribbon
16.91 17. 48 18.86
24.07 28.95 33.84
88 ± 10 88 ± 16 92 ± 16
59.31 71.34 87.43
161.88 178.22 374.48
988 ± 137 980 ± 76 1020 ± 235
surface is consistent with their volume fraction in the bulk, then the change in polarization ratio would be of the order 1%, which is at the limits of detection for the instrument used. In order to probe the bulk nanomorphology of the composite samples at length scales from ∼1 to 200 nm, we used SANS, and the results are shown in Figure 4. In SANS, the intensity is proportional to the number, size, and contrast of the scattering entities in a sample, while the q-dependence of the intensity is related to their shape and local arrangement. So in our composites it is the scattering from the C60 that dominates the SANS, not the hydrogenous PS (which would be expected to contribute a mostly flat background). We have started our analysis of the SANS data by determining if processing induces any nanoscale orientation in our samples. In the Supporting Information, Figure S.I.4.(a)
No structure was observed in the pre-solvent blended sample on SESANS length scales; however, some depolarization is observed ⟨P⟩ = 0.9953 which can be attributed to very small structures (well below the micrometer level observed in the mixed powder sample). In the micrographs obtained by optical microscopy there is some evidence for larger structures in the pre-solvent blended sample. However, Figure 2b shows that the area ratio is 1 order of magnitude smaller than that of the mixed powders. The change in SESANS polarization ratio as a function of spin-echo length is given by P = P0 exp(Σt[G(z) − 1]), where G(z) is the projection of the density distribution along the measurement axis and the term Σt (the fraction of neutrons that are scattered only once) in a binary mixture scales with the volume fraction Φ as Φ(1 − Φ). Hence, if one assumes that the area ratio of agglomerates measured at the E
DOI: 10.1021/acs.macromol.6b02283 Macromolecules XXXX, XXX, XXX−XXX
Article
Macromolecules
Figure 3. Spin-echo SANS signals, expressed as the depolarisation ratio P/P0, of extruded PS−C60 nanocomposites extracted from sampling port 9. The black line is a fit to a monodisperse sphere model while the blue line is a constant illustrating a small depolarisation over those length scales probed in the pre-solvent blended sample.
shows the 2D SANS patterns from extruded ribbons processed from both a pre-solvent blend and from mixed powders, and in Figure S.I.4.(b) we show the corresponding intensity ratios I(QV)/I(QH) as a function of q for the vertical and horizontal quadrants of the 2D patterns. As clearly shown, the nanoscale distribution of C60 aggregates in the extruded ribbons is isotropic. Accordingly, the SANS data were circularly averaged to produce 1D graphs of intensity vs q. As shown in Figure 4a for the pre-solvent blended composite, five different samples were analyzed using SANS, namely, (a) a compression-molded sample annealed at 90 °C for 5 min, (b) a similarly compression-molded sample annealed at 170 °C for 5 min, (c) an extruded sample collected at P2, (d) an extruded sample collected at P9, and (e) an extruded ribbon. As can be seen in Figure 4a, the neutron scattering intensity I(q) of the compression-molded sample annealed at 90 °C for 5 min is very weak and only increases slightly when the annealing temperature increases to 170 °C. Our goal on testing these samples at these two temperatures (90 and 170 °C), intermediate between room temperature and the extrusion temperature (200 °C), was to study the effect of temperature alone on the nanomorphology of the composite samples as they are heated at the very beginning of the extruder. When the composites are fed into the extruder, an abrupt increase of nearly 1 order of magnitude in scattering intensity occurs in the early stages of the extrusion process, as revealed
Figure 4. SANS data from extruded PS−C60 nanocomposites (identical scales have been used for ease of comparison): (a) presolvent blended; (b) prepared from mixed powders of PS and C60.
by a comparison between the two previous samples and the sample P2. Indeed, it can be seen that the scattering intensity continues to increase further along the extruder, reaching a maximum value in the extruded ribbon. Figure 4b presents the SANS data for the corresponding composites prepared from mixed powders of the pure components PS and C60. Also shown for comparison are the background (incoherent) SANS scattering of a ribbon of pure PS extruded under the same conditions as well as the SANS of a sample prepared by compression molding at 170 °C
Table 2. SANS Fitting Parameters (DAB Model) for the q Range 0.009−0.2545 Å−1 a CDB (Å−4) pre-solvent blended
from mixed powders
90 °C 170 °C P2 P9 ribbon P2 P9 ribbon
2.5 3.9 3.8 7.3 9.7 3.2 3.0 2.4
× × × × × × × ×
−6
10 10−6 10−5 10−5 10−5 10−5 10−5 10−5
± ± ± ± ± ± ± ±
2.9 1.2 8.4 1.1 2.2 3.8 3.9 7.6
Δρ (Å−2) × × × × × × × ×
−7
10 10−6 10−7 10−6 10−6 10−7 10−7 10−7
3.91 4.88 1.52 2.11 2.44 1.40 1.35 1.21
× × × × × × × ×
−7
10 10−7 10−6 10−6 10−6 10−6 10−6 10−6
length L (nm) 14.7 25.0 20.7 20.5 25.5 14.9 15.6 14.9
± ± ± ± ± ± ± ±
4.1 12.3 0.8 0.5 0.9 0.4 0.5 1.2
m 1.2 1.8 2.7 3.1 3.3 2.6 2.5 2.4
± ± ± ± ± ± ± ±
0.3 0.3 0.03 0.02 0.02 0.03 0.03 0.07
CDB is the scaling factor, Δρ is the contrast between the two phases, L is the average distance between the two phases, and m is the slope of the SANS data in their linear region extracted using the power law model (I(q) ∝ q‑m).
a
F
DOI: 10.1021/acs.macromol.6b02283 Macromolecules XXXX, XXX, XXX−XXX
Article
Macromolecules
results most likely from some large-scale heterogeneities (e.g., voids). The values obtained from the fitting for CDB and L using eq 1 are shown in Table 2. The values of Δρ calculated from the respective CDB values are also included. All these values are depicted graphically in Figure 5.
consisting of C60 and PS powders in the weight ratio 1:99. This sample corresponds to the unmixed reference, i.e., a sample in which C60 and PS are not intermixed (situation A in Figure 1). As seen in Figure 4b, there is again an almost 1 order of magnitude increase in scattering intensity in the early part of the extruder, i.e., from the powder mixture at the inlet (here represented by our “unmixed reference”) to sampling location P2. But, contrary to what is observed with the pre-solvent blended composites, the scattering intensity changes very little with processing from sampling location P2 to the die outlet. We continued our SANS analysis by determining the slopes of the SANS data in their linear regions (between 0.009 and 0.015 Å−1 for the pre-solvent processed composite samples compression molded at 90 and 170 °C and between 0.009 and 0.025 Å−1 for the pre-solvent processed composite samples collected from the extruder and for the composite samples prepared from mixed powders) with a simple power law model fit of the form I(q) ∝ q−m, where m is related to the fractal dimension of the underlying structure. Considering first the samples prepared from pre-solvent processed composites, we find the value of m increases from 1.2 at 90 °C, to 1.8 at 170 °C, and to 2.7, 3.1, and 3.3 at sampling locations P2, P9, and the extruded ribbon, respectively (see details in Table 2). This indicates that with increasing temperature and residence time inside the extruder, the mass distribution of C60 changes from something sparsely distributed to something more clustered, with the PS−C60 interfaces becoming better defined. These factors point to nanoagglomeration of the C60. Turning now to the composite samples prepared from mixed powders of PS and C60, we find that the value of m exhibits a small decrease from 2.6 in P2, 2.5 in P9, and 2.4 in the extruded ribbon (Table 2). This indicates that processing induces a slight shift in the C60 mass distribution toward something less clustered and with more diffuse PS−C60 interfaces, suggesting some improvement of dispersion in these composites. A more quantitative approach to interpreting the SANS data is to model-fit it over its full range (q = 0.009−0.2545 Å−1). The model we have chosen to use for this is the Debye− Bueche (DB), also called the Debye−Anderson−Brumberger (DAB), model44,45 (eq 1). This model calculates the scattering from a randomly distributed, two-phase system that is characterized by a single length scalethe correlation length, Lwhich is a measure of the average spacing between regions of the two different phases (1 and 2). Crucially, this model makes no assumptions about the underlying morphology of the sample. The DAB function has the form I(q) = C DB
L3 + background (1 + (qL)2 )2
Figure 5. Evolution of the correlation length (L), phase contrast (Δρ), and slope (m) with processing for the samples prepared: (a) from presolvent blended composites; (b) from mixed powders of pure PS and pure C60.
According to the values in Figure 5a and in Table 2, in the case of the samples prepared from solvent-processed blends, there is some increase in correlation length of the phase domains along the barrel of the extruder which indicates the occurrence of some phase domain coarsening. However, the most striking feature of these samples is the ∼1 order of magnitude increase in the scattering length density difference between the two phases (Δρ), from 3.91 × 10−7 Å−2 in the initial sample to 2.44 × 10−6 Å−2 in the extruded ribbon, which clearly indicates that during processing the phases become purer due to C60 reagglomeration. In the extreme situation of having pure phases of PS and pure phases of C60 the corresponding Δρ would be 4.19 × 10−6 Å−2. According to Figure 5b and Table 2, in the case of composite samples prepared from powder mixtures of pure C60 and PS, the correlation length between the two phases remains approximately constant (∼15 nm) from sample P2 to the extruded ribbon, and the phase contrast decreases slightly, suggesting only a slight improvement in the quality of the nanodispersion. In summary, therefore our SANS results show that the major changes in the nanomorphology of the PS−C60 composites occur in the early stages of the extrusion process, between the
(1)
where the prefactor CDB = 8πK(Δρ)2ϕ1ϕ2, where Δρ is the neutron scattering length density difference between the phases having volume fractions of ϕ1 and ϕ2 and K is a scalar to convert the units of L−1 to those of I(q) (i.e., cm−1). As shown in Figure 4, the DAB model (solid lines) gives a good description of the data. We have not fitted the data for the sample of mixed powders compression molded at 170 °C (sample “mixed powders-initial” in Figure S.I.2) and for the extruded ribbon of pure PS because neither of these samples conforms to the physics of the DAB model. The small amount of residual low-q scattering that is evident in these two samples G
DOI: 10.1021/acs.macromol.6b02283 Macromolecules XXXX, XXX, XXX−XXX
Article
Macromolecules
As seen, in both Figures 6a and 6b as well as in Figure S.I.5, the SAXS data show trends very similar to those observed in the corresponding SANS data, namely, (a) in the case of presolvent blended composites m increases progressively from m = 2.5 in the sample heated to 90 °C to m = 3.4 in the extruded ribbon; (b) in the case of composites from mixed powders m changes only very little between sampling location P2 and the extruded ribbon, specifically a small decrease is observed from m = 3.0 in location P2 to m = 2.9 in the extruded ribbon. These observations corroborate the previous SANS results showing that: (a) in the case of samples prepared from pre-solvent processed blends, PS−C60 interfaces become progressively sharper as the material progresses along the screw which points strongly to the occurrence of C60 reagglomeration; (b) in the case of samples prepared from mixed powders, there is a slight improvement in mixing from sampling location P2 to the extruded ribbon with the PS−C60 interfaces becoming slightly less sharp. The 1D WAXS patterns for the two samples prepared from the two different feeding formulations and collected from sampling port P9 are shown in Figure S.I.6. In both samples the two broad amorphous halos with maxima at q = 0.75 Å−1 and q = 1.35 Å−1 are due to the amorphous polystyrene. For mixed powders, at location P9 a crystalline C60 peak is clearly visible at q = 0.77 Å−1 as well as two small peaks at q = 1.26 and 1.48 Å−1.34 These peaks, which are not visible in the pre-solvent processed sample, are due to the presence of crystalline C60 aggregates.21,34 These results reinforce those obtained by OM and SESANS showing that extruded samples processed from mixed powders retain crystalline C60 aggregates while the samples processed from pre-solvent blend either do not have any crystalline C60 aggregates or have them in such a low quantity that they are below the detection limit of the WAXS technique. To investigate this morphological evolution further, we have performed some transmission electron microscopy (TEM) analysis of the samples collected from P2 and P9 for both the composites prepared from mixed powders and from pre-solvent blend. Figures S.I.7 and S.I.8 show some representative TEM images. The TEM images show that in the case of composites processed from mixed powders the size and number of aggregates decrease along the extruder, and in the case of composites processed from the solvent blend the size and number of the aggregates increase along the extruder, a finding in good qualitative agreement with our SANS data. Before proceeding further, we believe it is important to make a few comments about the complementarity of some of the different direct measurement techniques for assessing nanocomposite morphology. While TEM is arguably more accessible than neutron scattering, it nonetheless has some drawbacks, particularly for studying polymer composites. First, there is the issue of the representative nature of the sampling process: TEM is subjective in the sense that the operator must search the sample for regions of interest, and those regions of interest must be at or near to the surface of the sample. In contrast, SANS is a bulk sampling technique: the beam is essentially the same size as the sample and the weak neutron−nucleus interaction conveys depth penetration. In the present work we have used a 10 mm diameter neutron beam, meaning the illuminated area ∼78 mm2 (= 78 000 000 μm2). However, in the TEM pictures shown in the Supporting Information the maximum area probed in a single picture (Figure S.I.7.(a)) is only ∼60 μm2. SAXS suffers from a similar limitation to TEM
hopper and sampling port 2. This is in accordance with observations reported for other polymer-based systems, such as polymer blends and composites containing layered silicates, carbon nanotubes, or graphene derivatives. As the material reaches the first kneading zone of the screw, it melts as a result of a combination of conducted and dissipated heat. Since melt temperatures are still low, hydrodynamic stresses are high, favoring dispersion. Moreover, flow through kneading disks is complex, promoting distribution. Furthermore, according to our SANS results, in the case of composites prepared from the pre-solvent blend the main morphological evolution that occurs along the extruder is a continuous reagglomeration of the initially molecularly dispersed C60 into purer phase domains. In the case of composites prepared from mixed powders the main morphological evolution is an initial abrupt increase in the degree of dispersion up to sampling location P2, followed by a much less visible morphological evolution from P2 to the die exit. Figure 6 illustrates the SAXS data for the same samples as in Figure 4, namely in Figure 6a for the pre-solvent blended
Figure 6. Small-angle X-ray scattering of extruded PS and PS−C60 nanocomposites: (a) pre-solvent blended; (b) prepared from mixed powders of PS and C60.
composites heated to 90 and 170 °C, and also collected from P2, P9, and extruded ribbon, and in Figure 6b for the composites compounded from mixed powders and collected at P2, P9, and extruded ribbon. Again, data in the q range 0.009− 0.025 Å−1 were fitted with a linear power law and obtained values of m. These are plotted in Figure S.I.5. H
DOI: 10.1021/acs.macromol.6b02283 Macromolecules XXXX, XXX, XXX−XXX
Article
Macromolecules
solvent-blended composites (Figure S.I.9) which showed no clear evidence of its presence, within the resolution of the technique. However, we did observe contamination on surface of solvent processed samples caused by the electron beam (Figure S.I.10). This well-known surface contamination is electron beam induced deposition of excess volatile carbonaceous species. These observations strongly suggest that trapped-solvent is present in very small amounts, although enough to decrease the Tg of the system by several °C. The most interesting feature of the Tg of the solvent-blended composites is that it drops from ∼99.7 to ∼96.6 °C respectively at P2 and P9, and then it remains approximately constant between P9 and the extruded ribbon. Based also on previous work,19,32,34 this decrease in Tg along the extruder is most likely due to the fact that C60 is leaving the PS matrix and reagglomerating, as also supported by our SANS results. Figure 8 shows the mechanical spectra of the PS matrix and the composites obtained when processing the pre-solvent blended composites (a) or the mixed powders (b).
in this respect. Second, to achieve any penetration into a sample with electron or X-ray beams, it is necessary to use beams of high energy (many keV), energies that are way beyond covalent bond energies meaning that there is a significant possibility of radiation-induced damage in the sample. The “cold” neutrons used in SANS, on the other hand, have energies of just a few meV, meaning SANS is a genuinely nondestructive technique. Lastly, there is the issue of phase contrast. Electrons and X-rays interact with atomic electrons meaning that the greatest contrast is obtained between elements of significantly different atomic number. In our samples we have carbon dispersed in a matrix of carbon and hydrogen. This means that for example, in the TEM pictures we have provided in the Supporting Information, there is no reliable way of knowing if what we see are aggregates of C60 in a matrix of pure PS or regions rich in C60 in a soup of PS with some very well molecularly dispersed C60. In SANS, however, the phase contrast also stems from the neutron−nucleus interaction and can be very different between nuclei of similar atomic number, as it indeed is between carbon and hydrogen. Thus, while our neutron measurements do not provide the same visually intuitive picture of a sample that TEM does, they are sensitive to structural information that TEM simply is not. In Figure 7, we show the DSC data, where we plot the variation of the glass transition temperature Tg with processing.
Figure 8. Frequency dependence of the dynamic viscosity |η*| of the solvent prepared composites (a) and the mixed powders composites (b) sampled at different locations along the extruder. Lines correspond to the dynamic viscosity of the PS matrix fed into the extruder whereas dashed lines represent the dynamic viscosity of an extruded PS ribbon.
Figure 7. Variation of the glass transition temperature (Tg) along processing. Tg was determined using the extrapolated half-Cp method. Associated error bars are typically